Highly Accurate and Precise Automated Cup-to-Disc Ratio Quantification for Glaucoma Screening.

Objective: An enlarged cup-to-disc ratio (CDR) is a hallmark of glaucomatous optic neuropathy. Manual assessment of the CDR may be less accurate and more time-consuming than automated methods. Here, we sought to develop and validate a deep learning-based algorithm to automatically determine the CDR from fundus images. Design: Algorithm development for estimating CDR using fundus data from a population-based observational study. Participants: A total of 181 768 fundus images from the United Kingdom Biobank (UKBB), Drishti_GS, and EyePACS. Methods: FastAI and PyTorch libraries were used to train a convolutional neural network-based model on fundus images from the UKBB. Models were constructed to determine image gradability (classification analysis) as well as to estimate CDR (regression analysis). The best-performing model was then validated for use in glaucoma screening using a multiethnic dataset from EyePACS and Drishti_GS. Main Outcome Measures: The area under the receiver operating characteristic curve and coefficient of determination. Results: Our gradability model vgg19_batch normalization (bn) achieved an accuracy of 97.13% on a validation set of 16 045 images, with 99.26% precision and area under the receiver operating characteristic curve of 96.56%. Using regression analysis, our best-performing model (trained on the vgg19_bn architecture) attained a coefficient of determination of 0.8514 (95% confidence interval [CI]: 0.8459-0.8568), while the mean squared error was 0.0050 (95% CI: 0.0048-0.0051) and mean absolute error was 0.0551 (95% CI: 0.0543-0.0559) on a validation set of 12 183 images for determining CDR. The regression point was converted into classification metrics using a tolerance of 0.2 for 20 classes; the classification metrics achieved an accuracy of 99.20%. The EyePACS dataset (98 172 healthy, 3270 glaucoma) was then used to externally validate the model for glaucoma classification, with an accuracy, sensitivity, and specificity of 82.49%, 72.02%, and 82.83%, respectively. Conclusions: Our models were precise in determining image gradability and estimating CDR. Although our artificial intelligence-derived CDR estimates achieve high accuracy, the CDR threshold for glaucoma screening will vary depending on other clinical parameters. Financial Disclosures: Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

Assessing the Efficacy of Synthetic Optic Disc Images for Detecting Glaucomatous Optic Neuropathy Using Deep Learning.

Purpose: Deep learning architectures can automatically learn complex features and patterns associated with glaucomatous optic neuropathy (GON). However, developing robust algorithms requires a large number of data sets. We sought to train an adversarial model for generating high-quality optic disc images from a large, diverse data set and then assessed the performance of models on generated synthetic images for detecting GON. Methods: A total of 17,060 (6874 glaucomatous and 10,186 healthy) fundus images were used to train deep convolutional generative adversarial networks (DCGANs) for synthesizing disc images for both classes. We then trained two models to detect GON, one solely on these synthetic images and another on a mixed data set (synthetic and real clinical images). Both the models were externally validated on a data set not used for training. The multiple classification metrics were evaluated with 95% confidence intervals. Models' decision-making processes were assessed using gradient-weighted class activation mapping (Grad-CAM) techniques. Results: Following receiver operating characteristic curve analysis, an optimal cup-to-disc ratio threshold for detecting GON from the training data was found to be 0.619. DCGANs generated high-quality synthetic disc images for healthy and glaucomatous eyes. When trained on a mixed data set, the model's area under the receiver operating characteristic curve attained 99.85% on internal validation and 86.45% on external validation. Grad-CAM saliency maps were primarily centered on the optic nerve head, indicating a more precise and clinically relevant attention area of the fundus image. Conclusions: Although our model performed well on synthetic data, training on a mixed data set demonstrated better performance and generalization. Integrating synthetic and real clinical images can optimize the performance of a deep learning model in glaucoma detection. Translational Relevance: Optimizing deep learning models for glaucoma detection through integrating DCGAN-generated synthetic and real-world clinical data can be improved and generalized in clinical practice.

Diagnostic Accuracy of Artificial Intelligence in Glaucoma Screening and Clinical Practice.

PURPOSE: Artificial intelligence (AI) has been shown as a diagnostic tool for glaucoma detection through imaging modalities. However, these tools are yet to be deployed into clinical practice. This meta-analysis determined overall AI performance for glaucoma diagnosis and identified potential factors affecting their implementation. METHODS: We searched databases (Embase, Medline, Web of Science, and Scopus) for studies that developed or investigated the use of AI for glaucoma detection using fundus and optical coherence tomography (OCT) images. A bivariate random-effects model was used to determine the summary estimates for diagnostic outcomes. The Preferred Reporting Items for Systematic Reviews and Meta-Analysis of Diagnostic Test Accuracy (PRISMA-DTA) extension was followed, and the Quality Assessment of Diagnostic Accuracy Studies-2 (QUADAS-2) tool was used for bias and applicability assessment. RESULTS: Seventy-nine articles met inclusion criteria, with a subset of 66 containing adequate data for quantitative analysis. The pooled area under receiver operating characteristic curve across all studies for glaucoma detection was 96.3%, with a sensitivity of 92.0% (95% confidence interval: 89.0-94.0) and specificity of 94.0% (95% confidence interval: 92.0-95.0). The pooled area under receiver operating characteristic curve on fundus and OCT images was 96.2% and 96.0%, respectively. Mixed data set and external data validation had unsatisfactory diagnostic outcomes. CONCLUSION: Although AI has the potential to revolutionize glaucoma care, this meta-analysis highlights that before such algorithms can be implemented into clinical care, a number of issues need to be addressed. With substantial heterogeneity across studies, many factors were found to affect the diagnostic performance. We recommend implementing a standard diagnostic protocol for grading, implementing external data validation, and analysis across different ethnicity groups.

The Descemet Membrane in Primary Congenital Glaucoma.

PURPOSE: To evaluate Descemet membrane (DM) morphology in eyes with primary congenital glaucoma (PCG) in vivo using high-definition anterior segment optical coherence tomography (ASOCT) and on histopathology. METHODS: Corneal scans of patients with PCG (22 eyes of 15 patients) were evaluated for DM morphology and anterior chamber angle using ASOCT. The DM thickness in PCG eyes was compared with fellow eyes (8 eyes) of unilateral patients with PCG and healthy controls (12 eyes) on ASOCT. The DM morphology was also compared on the histopathology of corneal tissues (9) obtained from PCG eyes after keratoplasty and enucleated eyes of retinoblastoma (6 controls) on light microscopy with immunostaining for collagen IV. RESULTS: On ASOCT, all affected eyes showed the presence of either a thickened DM complex or a hyper-reflective double layer representing the thickened DM and pre-Descemet layer (PDL), unlike a single membrane in the controls and fellow eyes. On ASOCT, among patients with PCG, the DM showed significant thickening (32.0 ± 11.2 µm) versus fellow eyes (14.4 ± 3.3 µm) and controls (11.5 ± 1 µm) (P < 0.001; analysis of variance). The thickened DM complex continued peripherally into the trabecular meshwork as an abnormal membrane in 16/22 affected eyes. On histopathology, thickening of DM was also more among PCG eyes (median: 67.9 µm range: 27.2-214.9) versus controls (median: 27.7 µm, range: 22.1-36.1; P = 0.005) as also of PDL (median: 14 µm, range: 5.9-30.5) of PCG versus (median 3.5, range: 1.3-6.7 µm) in controls; P = 0.014. CONCLUSIONS: Thickening of DM and PDL occurs in eyes with PCG and is seen to have a peripheral extension upto the angle recess.

Short-term incidence and management of glaucoma after successful surgery for stage 4 retinopathy of prematurity.

Purpose: The purpose of this study is to describe the short-term incidence, clinical features, and management of glaucoma in children after successful surgery for stage 4 retinopathy of prematurity (ROP). Methods: The retrospective study included all eyes undergoing successful surgery for stage 4 ROP with good outcomes at a tertiary eye care center between June 2014 and June 2016. Cases developing postoperative glaucoma underwent examination under anesthesia for measurement of intraocular pressures (IOP), corneal diameters, Retcam-assisted fundus imaging, and gonioscopy. Outcomes of glaucoma management were evaluated. Results: Hundred eyes of 70 babies underwent successful surgery for stage 4 ROP (with postoperative attached retina, and minimal sequelae) with minimum follow-up of 15 months. Six eyes (6%) developed postoperative glaucoma. Of these, four eyes had undergone lens-sparing vitrectomy and two were managed with lensectomy and vitrectomy (LV). Median time duration for development of glaucoma after primary vitreous surgery was 17.5 weeks. Two cases could be managed with topical IOP-lowering agents alone, whereas four required filtering surgeries (trabeculotomy with trabeculectomy and 0.04% mitomycin C [MMC] application). Average IOP decreased from 25 ± 2.36 to 12.2 ± 2.05 mmHg at 12 months from glaucoma diagnosis. Conclusion: Glaucoma is a potential adverse event following successful vitreous surgery for stage 4 ROP. A combined trabeculotomy-trabeculectomy along with MMC gives favorable outcome.

Comparison of Rates of Fast and Catastrophic Visual Field Loss in Three Glaucoma Subtypes.

Purpose: To compare the distribution of visual field progression rates in three subgroups of glaucoma, being primary angle-closure glaucoma (PACG), POAG, and juvenile open-angle glaucoma (JOAG). Methods: We assessed glaucoma patients treated in an Indian tertiary care setting with at least four visual field assessments. We determined rates from a single eye of each of 525 patients using linear regression of the summary index mean deviation (MD) over time. The main outcome measures were the proportions of fast (<-1.0 to -2.0 dB/y) and catastrophic (<-2 dB/y) visual field progression. Bootstrapped 95% CIs allowed comparison with published data from a large clinical cohort in Canada. Results: The combined proportion of fast and catastrophic progressors in our cohort was less than half that in the Canada dataset (2.3% vs. 5.8%), despite median progression rates differing by only 0.03 dB/y. PACG, POAG, and JOAG represented 45%, 32%, and 12% of our cohort, respectively. Baseline MD values were similarly distributed between these subtypes. All subtypes showed a similar shaped distribution for progression rates, with median progression rates of -0.03, -0.05, and 0.02 dB/y for PACG, POAG, and JOAG, respectively. Combined proportions of fast and catastrophic progression rates did not significantly differ between subtypes. Conclusions: Differences in fast and catastrophic visual field progression can exist despite only small changes in median progression rates, highlighting the importance of considering the full shape of the progression rate distribution when comparing the risk of devastating visual field loss.

Visual Disability Among Juvenile Open-angle Glaucoma Patients.

AIM: Juvenile onset primary open-angle glaucoma (JOAG) unlike adult onset primary open-angle glaucoma presents with high intraocular pressure and diffuse visual field loss, which if left untreated leads to severe visual disability. The study aimed to evaluate the extent of visual disability among JOAG patients presenting to a tertiary eye care facility. METHODS: Visual acuity and perimetry records of unrelated JOAG patients presenting to our Glaucoma facility were analyzed. Low vision and blindness was categorized by the WHO criteria and percentage impairment was calculated as per the guidelines provided by the American Medical Association (AMA). RESULTS: Fifty-two (15%) of the 348 JOAG patients were bilaterally blind at presentation and 32 (9%) had low vision according to WHO criteria. Ninety JOAG patients (26%) had a visual impairment of 75% or more. CONCLUSIONS: Visual disability at presentation among JOAG patients is high. This entails a huge economic burden, given their young age and associated social responsibilities.

Role of CYP1B1, p.E229K and p.R368H mutations among 120 families with sporadic juvenile onset open-angle glaucoma.

BACKGROUND: To determine the frequency of CYP1B1 p.E229K and p.R368H, gene mutations in a cohort of sporadic juvenile onset open-angle glaucoma (JOAG) patients and to evaluate their genotype/phenotype correlation. METHODS: Unrelated JOAG patients whose first-degree relatives had been examined and found to be unaffected were included in the study. The patients and their parents were screened for p.E229K and p.R368H mutations. The phenotypic characteristics were compared between probands carrying the mutations and those who did not carry these mutations. RESULTS: Out of 120 JOAG patients included in the study, the p.E229K mutation was seen in 9 probands (7.5%) and p.R368H in 7 (5.8%). The average age of onset of the disease (p = 0.3) and the highest untreated IOP (p = 0.4) among those carrying mutations was not significantly different from those who did not have these mutations. The proportion of probands with angle dysgenesis among those with p.E229K and p.R368H mutations was 70% (11 out of 16) in comparison to 65% (67 out of 104) of those who did not harbour these mutations (p = 0.56). Similarly, the probands with moderate to high myopia among those with p.E229K and p.R368H mutations was 20% (3 out of 16) in comparison to 18% (18 out of 104) of those who did not harbour these mutations (p = 0.59). CONCLUSION: The frequency of p.E229K and p.R368H mutations of the CYP1B1 gene is low even among sporadic JOAG patients. Moreover, there is no clinical correlation between the presence of these mutations and disease severity.

In Vivo Analysis of Angle Dysgenesis in Primary Congenital, Juvenile, and Adult-Onset Open Angle Glaucoma.

Purpose: The purpose of this study was to comparatively evaluate angle dysgenesis in vivo, among congenital, juvenile, and adult-onset open angle glaucoma patients. Methods: A cross-sectional evaluation of 96 glaucoma patients, 22 children with primary congenital glaucoma (PCG) old enough to cooperate for optical coherence tomography (OCT), 34 juvenile-onset open angle glaucoma (JOAG) patients, 40 adult-onset primary open angle glaucoma (POAG), and 30 healthy subjects, was carried out using high-resolution anterior segment spectral domain (SD)-OCT. Subgroup analysis was done for presence/ absence of angle dysgenesis as defined by presence of abnormal tissue/hyperreflective membrane within angle recess and/or absence of Schlemm's canal (SC). Results: Morphologic features suggestive of angle dysgenesis such as the presence of abnormal tissue at the angle and a hyperreflective membranous structure covering the meshwork were seen in all PCG eyes (100%), in 14 (40%) JOAG eyes, and none of the POAG eyes in comparison to healthy eyes (P = 0.01, P = 0.03, and P = 0.23 for PCG, JOAG, and POAG, respectively). SC could be seen in 27 (90%) healthy eyes compared with only 7 (30%) in PCG (P = 0.01) 20 (60%) JOAG eyes (P = 0.03), and 26 (65%) adult-onset POAG eyes (P = 0.23; χ2 test). Conclusions: Angle dysgenesis in the form of abnormal tissue at the angle/hyperreflective membrane and/or absence of SC could be identified on anterior segment SD-OCT, which can be used for in vivo evaluation of eyes with developmental glaucoma.

Long-term outcomes of glaucoma drainage devices for glaucoma post-vitreoretinal surgery with silicone oil insertion: a prospective evaluation.

PURPOSE: To evaluate long-term success of the Ahmed glaucoma valve (AGV) for refractory glaucoma after vitreoretinal surgery with silicone oil insertion. METHODS: Prospective non-comparative evaluation of patients who underwent AGV insertion for management of post-vitreoretinal surgery glaucoma, post-silicone oil removal. Intraocular pressure (IOP), visual acuity, and glaucomatous neuropathy status were evaluated preoperatively and at multiple follow-up visits postoperatively. Success, using Kaplan-Meier analysis, was determined at the 12-month follow-up visit and at the last follow-up. Factors associated with failure were analysed. RESULTS: Twenty-seven eyes of 27 patients with a mean age of 28.3 ± 15.2 years underwent a superior AGV implantation. The average follow-up after AGV implantation was 17.11 ± 8.36 months (range: 9-60 months). Kaplan-Meier survival analysis revealed a 62 % success at 12 months and 37 % at 5 years. A 48 % rate of complications was noted, 22 % of which were vision-threatening. Factors analysed, including patient age, interval between vitreoretinal surgery and silicone oil removal, interval between vitreoretinal surgery and AGV implantation, and phakic status, were not found to be associated with higher failure rates. CONCLUSION: Long-term success of AGV implantation for glaucoma after vitreoretinal surgery with silicone oil insertion is better than that reported for trabeculectomy, though complication rates remain high.